I want you to picture something. You own a Ferrari. It's this marvel of engineering, incredible speed, precision, just raw power. Right. But every single day, you only use it to drive two blocks to the grocery store. That is the perfect analogy. I mean, we have this engine of like infinite potential sitting right on our desktops, and we're just using it to write out of office emails. Today, we are going to stop driving to the grocery store. Welcome to the Deep Dive. It is January 21st,
2026. And we are getting an early look at a guide dropping tomorrow. It's called 12 Advanced Chat GPT Tricks for Power Users. And it's basically the manual for the Ferrari that nobody ever bothered to read. The premise here is interesting. The author argues there's this huge gap between the tourist and the wizard. The tourist asks for a summary. The wizard, well, what is the wizard doing differently? Oh, the wizard is playing a completely different game. They're building
memory systems. They have incognito modes for their thoughts. They have AI agents building software for them, you know, while they sleep. So let's bridge that gap. We have 12 tricks to get through today. Let's start with section one, which is really about the mindset. The guide says that even now, in 2026, something like 95 % of users are missing these hidden features.
They're still acting like it's 2023. Right. If you're still typing, summarize this or write me a blog post, you're essentially eating AI fast food. The system can do it, sure, but there's no depth. A power user knows the tool isn't just a chatbot anymore. It's more like an operating system for your work. It makes you wonder, though. We have this tool that can simulate complex reasoning, yet we just default to the easiest path. Why do we settle for surface -level use? We simply
don't know what we don't know yet. Fair point. Okay, let's get into the mechanics. Trick number one is context -specific writing memory. This one solves the generic voice problem. So most people have one setting, one voice for ChatGPT, but you don't talk to your boss the same way you talk to your team. Hopefully not. That could get awkward. Exactly. So the trick here is you stop using a generic instruction like be professional.
Instead, you feed the AI actual examples. You paste in an email you wrote and say, analyze this for tone and style. And here's the key part. You then say, store this as team emails. So instead of prompting it every time, it just knows. If I say I'm writing to the team, it's casual. If it's for a client, it's polished. Yes. You build out these personas, a client persona that's super formal, a LinkedIn persona that's more punchy. The AI just remembers which mask to wear depending
on the context. It does raise a question about authenticity. If I have an algorithm managing my personas, does this erase our authentic voice? No, it just scales our different personas efficiently. Okay, trick number two shifts from how we speak to how we listen. It's called scheduled daily reports. Yeah, this is where we see the big shift from reactive to proactive. I mean, usually you have to go to ChatGPT and ask it for something. This lets the AI nudge you. So we're inviting
the machine to nag us now. In a useful way. You set a trigger. Every morning at 8 a .m., send me updates on OpenAI, Google, and Anthropic, and bam, a desktop alert. Or you could have it send you a weekly goal review every Friday afternoon. It's interesting. We spend so much of our digital lives trying to filter out notifications, and now we're programming a machine to send us more of them. Does this reduce the noise or just add to it? It filters the noise so you focus on signal.
Let's move to trick number three, project -only memory. This sounds like a privacy feature. It is, but it's also a sanity feature. By default, ChatGPT remembers everything account -wide. So if you tell it about your dog's birthday in one chat, it technically knows that when you're writing a legal brief in another. Which is not great, especially if you're juggling multiple clients. Right. So project -only memory is a toggle. You find it in the gear icon when you start a new
project, and it builds a walled garden. That project can only see the files and chats inside of it. Nothing else. So my data for client A and client B will never, ever cross -pollinate. Precisely. And the guide points out something really critical. You have to make this choice at the start. You can't change it later. It seems like we're constantly negotiating these boundaries with our tools. Is privacy the ultimate luxury in AI? Compartmentalization is safety. Leakage
is unprofessional. Moving on. Trick number four. New personalization settings. This is the death of the old text box. You know, we used to have to type out, please be professional, or explain this like I'm five. Now, under settings, there are just drop -down presets. Presets, like filters on a photo. Exactly like that. Friendly, candid, professional, you just click. But the real power move is stacking them. You choose the friendly preset. Then you fill out the about you section
with your specific goals. And then you use those memory tricks we just talked about. It's all about layering context. It's about alignment. The tool stops feeling like a generic robot and starts feeling like a colleague who actually gets you. I find those labels so interesting. Friendly. Why do we feel this need to anthropomorphize the tool with friendliness? Because we work better when the vibe matches ours. All right. Trick number five. Temporary chats. This is basically
the incognito mode for AI. There's a little icon in the top corner, a chat bubble with a slash through it. I see. And this does what exactly? When you click it, nothing is saved. No history, no memory is trained on it. It's just gone when you close the tab. What's the use case for that? Well, hypotheticals. Let's say you're a CEO. You might want to ask, if I were a competitor to my own company, How would I put us out of business? You do not want that stored in your
long -term account memory. Right. Or if you just need to do a quick one -off math problem. Exactly. You don't want a random question about baking soda to influence your coding advice three weeks later. It suggests that memory can be a burden. So why is forgetting just as important as remembering? Freedom to experiment without polluting the long -term data. Trick number six. Company knowledge base. Now we're getting into the enterprise stuff. Yeah, this is for the business plan users. It
connects ChatGPT to everything. Your Google Drive, your Slack, your HubSpot, everything. So it's not just searching the public web. It's searching our internal world. Right. You can ask, what meetings do I have tomorrow? And can you find the related files in Dropbox? It'll scan your calendar, find the meeting, go into Dropbox, find the right PDF, and summarize it for you. That used to be a 20 -minute scramble. Where is that file? Was it in Slack? Was it in an email?
Now it's one prompt. It turns the AI into a single search guard for your entire work life. It's efficient for sure. But are we outsourcing our organizational brain? We're reclaiming time spent hunting for files. Next up, trick number seven, chat GPT apps. This is the ecosystem play. You know how you tag someone on social media with the at tip symbol? You can now do that with software right inside the chat. Give me an example. How does that work? So you type at Photoshop in the
chat bar. You say, create a social media graphic for this post. It doesn't just make a flat image. It connects to Photoshop. And you can even edit the properties. Or at Notion to save your chat notes directly to a workspace. The lines between the apps are just dissolving. It's all becoming conversation. The chat bar is becoming the universal interface. You don't open apps anymore. You just call them when you need them. So is the interface itself just dissolving into conversation? Yes.
The chat bar is the new operating system. Trick number eight, updating custom GPTs. This is a classic maintenance issue. You might have built an amazing custom GPT six months ago when it was running on, say, GPT -4. But now we have GPT 5 .2. And for anyone who doesn't know, a custom GPT is basically a version of ChatGPT you've trained for a specific task. Correct. But here's the catch. They don't always update themselves. You have to physically go in, click
Edit, and manually select the newest model. Otherwise, you've got that Ferrari, but with a lawnmower engine inside. And you can even do this with GPTs that other people have built. Yep, there's a drop -down. If you're using a public one that feels a bit sluggish, check the model. You can often override it to the latest version yourself. It's like digital hygiene. Why is maintenance so often ignored in our digital tools? Invisible decay is easy to miss until performance drops.
Trick number nine, batch image generation. This one is all about speed, pure speed. DLE is now, what, 4x faster in this 2026 update? So you don't ask for one logo idea, you ask for four at once. Minimalist, bold, abstract, and gradient. And it generates them all simultaneously. All at once. It completely changes the creative workflow. It's not... Wait, critique anymore. It's generate a batch, pick the best one, and refine from there. But does that speed devalue the creative choice?
Speed allows for rapid iteration and better final choices. Now, trick number 10, canvas mode mini apps. I have to admit, when I was reading the source material, this is the one that really stopped me in my tracks. Oh, this is the wow moment. This is the one that really makes you feel like a wizard. Whoa. Just to define it for everyone listening, Canvas mode is a split screen that shows you code on one side and a live app on the other. Exactly. And you don't need to
know a single line of code. You can just say, create a to -do list with a dark mode, toggle, and local storage. And it just builds it. It just appears. It appears. Code on the left, a fully working app on the right. You can click the buttons. It works. You can build mortgage calculators, little games, productivity timers. And then you can just download the code and put it on your website. That's profound. We used to need a team of engineers for that. Now all
you need is a sentence. What happens when creation has zero friction? Everyone becomes a builder. Ideas become reality instantly. Trick number 11. Deep research plus custom knowledge. Okay, so... Deep research has been around for a little while. It scans the web. It reads articles. But the real trick is the combination. You upload your own files first. So you're anchoring the research in your own context. Yes. Let's say
you're researching market pricing. You upload your own internal pricing model spreadsheet. Then you say research competitor pricing. It doesn't just give you a generic list. It compares the world's data directly against your data. It's synthesizing the external with the internal. Context is king. Right. It stops giving you generic advice because now it knows your specific constraints. So is this the end of generic advice? Context
is king. Generic answers are now obsolete. And finally, trick number 12, extended thinking mode. This is one I struggle with personally. Oh, so? And I think a lot of us do this. I get lazy. I have this bad habit of just leaving the model on auto. I'll ask it a really complex question, get a mediocre answer, and then I spend the next 20 minutes trying to fix it. I call it prompt drift. I think we've all been there. I know I
have. You just want the answer now, so you try to coax intelligence out of the faster mode. Right. The trick is to manually select extended thinking mode. It forces the AI to pause. It actually thinks for, you know, 30 seconds to two minutes. It plans, and it checks its work before it gives you an answer. The difference between a reflex and a reflection. Exactly. For strategy, for architecture. For any complex problem,
you have to let it think. There's even a heavy mode for pro users that thinks for even longer. We're so used to instant gratification. Is patience the new skill we need to relearn? Waiting 30 seconds beats an hour of correcting errors. And there's a bonus trick in the guide, chat branching. This is for all the non -linear thinkers out there. You're in a conversation and you get to a fork in the road. Should I write a funny email or a serious one? And usually you have to pick
a path. Not anymore. You click. branch into new chat and suddenly you have two parallel timelines so just so we're clear branting means creating a new conversation path from an existing message right in one timeline you explore the funny email in the other you explore the serious one you can explore both realities at the same time without losing your original place it's like a multiverse of thought and you can save these branches into a project to synthesize them later it's like
a whiteboard strategy session but in text form Does this finally reflect how the human mind actually works? Absolutely. Nonlinear thinking is finally supported by the tool. Mid -role sponsor placeholder. So let's just try to unpack all of this. We've covered 12 tricks. Memory injection, proactive scheduling, privacy walls, app integrations, these deep thinking modes. It's a lot. It can definitely feel overwhelming. It is. But if you zoom out and look at the big picture. The theme
isn't really about the software. The software is just the engine. The real theme is about the driver. Exactly. The tool didn't get smarter between yesterday and today. You get smarter by knowing these buttons exist. We started with that Ferrari analogy. The difference between the person going to the grocery store and the race car driver isn't the car itself. It's the
knowledge of how to shift the gears. And knowing that you don't have to settle for the default settings, you can mold this thing to be your perfect co -pilot. So here's my challenge to you as you're listening right now. Don't try to do all 12 of these today. You'll just burn out. Pick one. Just one. Maybe it's setting up the about you personalization. Maybe it's trying that canvas mode to build a simple calculator. Yeah, just try one thing. See how it changes
the friction in your day -to -day workflow. The gap between the tourist and the wizard isn't intelligence. It's just curiosity. Go find the hidden buttons. See you in the deep end.